Two Technologies, One Old Question: What Makes Learning Happen?

A collaboratively authored post by Kelly Haylock and Carla Voyageur.

Claims about technology transforming education are everywhere, especially now that generative AI has entered nearly every corner of the classroom. In 1994, Richard Clark and Robert Kozma debated a version of the same question. Clark argued that media are only vehicles for delivering instruction, and that it is the instructional method, not the technology, that causes learning. Kozma reframed the question, asking not whether media influence learning but under what conditions they will, depending on a medium’s specific capabilities and how they interact with learners.

More than thirty years later, that debate is still the sharpest tool available for testing claims made about ed-tech. In this post, we examine two recent examples: a company blog post from Hurix Digital on AI tutoring, and a newspaper article on a virtual reality project for Indigenous health education. Using Clark and Kozma as two different lenses, we ask whether the benefits each claim come from the technology itself, from the instructional design behind it, or from both together.

Article 1: “How AI Tutors Personalize Education at Scale” (Hurix Digital, 2026)

This is a company blog post by Reena Shah, Vice President of Delivery at Hurix Digital, which sells digital content and learning services (Shah, 2026). The post argues that AI tutoring solves Bloom’s “two sigma problem,” the finding that one-on-one tutoring far outperforms classroom teaching but can’t be scaled. It says that in several peer-reviewed studies, AI tutoring outperforms one-on-one instruction, and it closes with an invitation to book a consultation.

The evidence does not match the claim. The Harvard trial it relies on compared and AI tutor with in-class active learning, not with a human tutor (Kestin et al., 2025). The second study, as the post describes it, supports a human-AI hybrid rather than AI alone. Market growth figures and student attitude surveys are offered as if they showed learning. In fairness, the post also calls AI tutoring “the delivery mechanism”.

Clark’s Response

He would find that word revealing. Calling AI a delivery mechanism concedes that method, not the medium, influences learning (Clark, 1994, pp. 23, 26). His replaceability test (p. 22) applies: Bloom’s result came from human tutors using mastery learning and feedback, a method humans can also deliver. What AI changes is cost and reach, which is a real but different claim. The Harvard authors themselves credit personalized feedback and self-pacing, both methods (Kestin et al., 2025). Market projections and attitude surveys measure enthusiasm, not learning (Clark, 1994, p. 27).

Kozma’s Response

He would give the post more credit. Immediate individual feedback and self-pacing are capabilities a classroom can’t easily offer, and he argues that capabilities enable methods (Kozma, 1994, p.16). But he would ask for the mechanism and process evidence (pp. 14-15). The study reports test scores, time on task, and ratings, but as far as we read, no analysis of the tutoring conversations. He would also ask for which learners and tasks, since the study covered two lessons in one physics course (Kestin et al., 2025). He warns that a student working alone with a computer recalls the teaching machine, and that media help only when designed into social context (pp. 16-17). The post’s advice to keep humans for motivation fits that view, but its headline claim that AI tutoring itself transforms learning does not.

Article 2: “U of S researchers use VR to revolutionize Indigenous health education” (The Saskatoon StarPhoenix, 2025)

This article, written by Aidan Jaager for The Saskatoon StarPhoenix, focuses on the VINE (Virtual Reality for Indigenous Youth and the Next Generation of Emerging Health Professionals) project, which uses VR to create more culturally responsive health sciences education resources and expand learning opportunities for Indigenous students. The VINE project uses 3D brain models that bring together Western neuroanatomy with traditional Indigenous teachings, along with 360-degree virtual campus tours for remote youth. VR is framed as a “modern technology avenue” to “revolutionize Indigenous health education,” positioning the technology as a major driver for addressing learning gaps and overcoming institutional barriers.

The project seeks to address curriculum gaps stemming from the historical exclusion of traditional Indigenous knowledge and community perspectives in standard healthcare education. It also addresses systemic and geographic barriers by tackling physical isolation in remote communities and the ongoing underrepresentation of Indigenous students in the health professions. Techno-determinism appears in the media headlines and narrative, which attributes systemic solutions directly to the medium, framing VR as a “modern technology avenue” inherently capable of “revolutionizing” education. At the same time, the project’s implementation resists pure determinism by centering a “community-driven, community-based” model led by Elders, Knowledge Keepers, and youth (Jaager, 2025). This tension highlights the divide between viewing technology as the primary driver of educational change and viewing it as a tool whose impact depends on how it is designed, implemented, and situated within a particular social and educational context.

The underlying research presents a more complicated picture than the media framing alone suggests. The NFRF project description describes VR as having “transformative potential” but positions that potential within a community-based participatory research process involving Indigenous youth and community members in co-developing the resources (NFRF, 2024)

Clark’s Response

Clark would separate the medium from the instructional conditions surrounding it. VR may be the delivery method, but the learning comes from the curriculum, Indigenous knowledge, community involvement, and student engagement (Clark, 1994, pp. 22–23). USask described the project as a “proof-of-concept,” that will hopefully lead to continued development (Olson, 2025). This matters for Clark because any eventual learning outcomes would raise the question of what can be attributed to VR itself and what comes from the curriculum, culturally relevant content, and community-based design surrounding it (Olson, 2025). His replaceability argument raises the question of whether someone could achieve the same learning through a home computer, video, or physical materials; Clark would question whether VR itself makes the difference (Clark, 1994, p. 22). The cost and access requirements also raise questions about whether its observed benefits come from the technology itself or reflect the additional resources and support it brings (Clark, 1994, p. 27).

Kozma’s Response

Kozma would consider how VR’s particular affordances shape the learning experience. Three-dimensional manipulation and 360° environments may enable forms of spatial exploration and interaction that are difficult to reproduce through a textbook or flat screen (Kozma, 1994, p. 11). This also supports the project’s stated rationale for using VR. The planned brain model allows students to move and manipulate a three-dimensional brain, which the researchers argue may help them understand the complexity of brain anatomy and neurophysiology while connecting Western health research with traditional Indigenous teachings (Olson, 2025). The Kozma question then becomes whether these specific capabilities actually contribute to learning and under what conditions. These capabilities operate within a social and instructional context; their value depends on how they are designed and used (Kozma, 1994, pp. 16–17). Community involvement, including Elders, youth, and Knowledge Keepers, helps shape what learners encounter and how they interpret the experience.

Conclusion

Read side by side; the Hurix and U of S articles show two different faces of techno-deterministic thinking. Hurix’s claim that AI tutoring outperforms one-on-one instruction rests on evidence that does not quite say that, and its framing of AI as a delivery mechanism supports Clark’s argument more than its own headline does. The U of S coverage is more hedged and community-led, but its opening claim that VR will revolutionize Indigenous health education still credits the technology before any results exist. In both cases, Clark would ask what instructional method is doing the work and whether it could be delivered another way, while Kozma would ask what capability of the medium is actually in play and whether it’s been designed into the right social and cultural context.

This is not only an academic exercise. In nursing, each new electronic medical record has arrived with the same promise: faster charting, easier communication, more time for patients. In practice, nurses now spend more time charting and less time at the bedside than before paper records disappeared, and when a system goes down, staff fall back on paper, the very method the technology was meant to replace. The EMR didn’t fail to deliver better care because the software was flawed; it succeeded at what it was actually built to do, capture more data, while the promise of more time with patients depended on staffing and workflow decisions the software was never designed to touch.

We leave readers with a simple habit: whenever a headline says a technology will transform learning, ask what method it is inside, and ask what evidence shows the medium itself, and not just its promise, is doing the work.

Transparency Statement

Generative AI was used to help analyze sources and refine wording: both authors verified citations and are responsible for final content.

References:

Clark, R. E. (1994). Media will never influence learning. Educational Technology Research and Development, 42(2), 21-29. https://www.jstor.org/stable/30218684

Jaager, A. (2025, July 29). U of S researchers use VR to revolutionize Indigenous health education. Saskatoon StarPhoenix. https://thestarphoenix.com/news/saskatchewan-news/u-of-s-researchers-use-vr-to-revolutionize-indigenous-health-education/

Kestin, G., Miller, K., Klales, A. et al. AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting. Sci Rep 15, 17458 (2025). https://doi.org/10.1038/s41598-025-97652-6

Kozma, R. B. (1994). Will Media Influence Learning? Reframing the Debate. Educational Technology Research and Development, 42(2), 7–19. http://www.jstor.org/stable/30218683

Olson, M. (2025, July 15). USask community research uses virtual reality for Indigenous health education. University of Saskatchewan. https://news.usask.ca/media-release-pages/2025/usask-community-research-uses-virtual-reality-for-indigenous-health-education.php

Shah, R. (2026, July 3). How AI tutors personalize education at scale [Blog post]. Hurix Digital. https://www.hurix.com/blogs/how-ai-tutors-personalize-education-at-scale/

Social Sciences and Humanities Research Council of Canada. (2024). Co-creating a VINE: Virtual Reality for Indigenous Youth and the Next Generation of Emerging Health Professionals. New Frontiers in Research Fund. https://sshrc-crsh.canada.ca/funding-financement/nfrf-fnfr/exploration/2024/award_recipients-titulaires_subvention-eng.aspx